Investigating the Effect of Anitpiracy Actions on Online Piracy

13-th IEEE International Conference on Peer-to-Peer Computing
Investigating the Reaction of BitTorrent Content
Publishers to Antipiracy Actions
∗
Reza Farahbakhsh∗ , Ángel Cuevas†∗ , Rubén Cuevas† , Reza Rejaie‡ ,
Michal Kryczka§ , Roberto Gonzalez† and Noel Crespi∗
Institut-Mines Télécom, Télécom SudParis, {reza.farahbakhsh, noel.crespi}@it-sudparis.eu
† Universidad Carlos III de Madrid {acrumin, rcuevas, rgonza1}@it.uc3m.es
‡ University of Oregon {[email protected]}
§ Institute IMDEA Networks {[email protected]}
Abstract—During recent years, a few countries have put in
place online antipiracy laws and there has been some major
enforcement actions against violators. This raises the question
that to what extent antipiracy actions have been effective in
deterring online piracy? This is a challenging issue to explore
because of the difficulty to capture user behavior, and to identify
the subtle effect of various underlying (and potentially opposing)
causes. In this paper, we tackle this question by examining
the impact of two major antipiracy actions, the closure of
Megaupload and the implementation of the French antipiracy
law, on publishers in the largest BitTorrent portal who are major
providers of copyrighted content online. We capture snapshots
of BitTorrent publishers at proper times relative to the targeted
antipiracy event and use the trends in the number and the level
of activity of these publishers to assess their reaction to these
events. Our investigation illustrates the importance of examining
the impact of antipiracy events on different groups of publishers
and provides valuable insights on the effect of selected major
antipiracy actions on publishers’ behavior.
Index Terms—Piracy, law, Cyberlocker, Megaupload, P2P,
BitTorrent, Hadopi.
I. I NTRODUCTION
During the past decade, the Internet has witnessed an
increasing level of online piracy of copyrighted content. In
particular, Peer-to-Peer content distribution applications (e.g.,
BitTorrent, Gnutella) and Cyberlocker services (e.g., Megaupload) have facilitated illegal sharing of copyrighted content1 .
At the same time, the availability of copyrighted content by
these systems at no cost, has led to an explosion in their
popularity and therefore to their contribution in overall Internet
traffic. While legal actions were taken against few major
and many minor violators who illegally published, consumed
or facilitated the distribution of copyrighted content, online
piracy appears to become even more widespread in different
countries. In recent years, these trends have prompted copyright holders to demand the legislation and implementation
of more effective online antipiracy laws in several countries.
However, such an effort has faced strong opposition by various
1 We want to make clear that Peer-to-Peer technology as well as Cyberlocker
facilities can be used for many other purposes further than sharing copyrighted
material. Furthermore, when we refer to illegal activity in this paper we
are not offering our opinion but just relaying on the fact that there are
court sentences and open processes using this term related to publishers of
copyrighted content. This paper does not include the opinion of the authors
regarding whether sharing copyrighted content is a legitimate action or not.
c
978-1-4799-0521-8/13/$31.00 ⃝2013
IEEE
stake holders in several countries. In fact, we are only aware
of a small number of countries that have legislated and
implemented an online antipiracy law. Given the difficulty to
put in place an online antipiracy law, an interesting question
is “whether and to what extent an antipiracy law and its
associated enforcement actions can affect the behavior of
violating users?”
This intriguing question is very difficult to answer for at
least three reasons as follows: First, the effect of an antipiracy
event (e.g., publicizing relevant laws or enforcement actions)
can be assessed on different groups of users including those
who publish or consume copyrighted content, users with different levels of involvement (as publisher or consumer), or users
for a specific system or in a particular country. Clearly, the
impact of an antipiracy event could vary significantly across
different groups. Second, there could be other (potentially
some unknown) co-existing social, economical, and technical
factors that have a dominant and possibly opposing effect on
piracy behavior among users. More importantly, it is very challenging to identify and capture all the relevant major factors,
and assess their level of impact on piracy behavior among
users. For example, the drop in the number of online pirating
for movies in the the US could be due to a combination of
antipiracy actions against a few users and/or due to user access
to cheap and legal content via Netflix. Furthermore, the effect
of an antipiracy action could be short- or long-lived. Third,
there is no ground-truth to reliably validate any finding about
user reactions to antipiracy events. A survey of users can be
conducted to obtain a more accurate view of the behavior for a
relatively small group of users (e.g., few thousands). However,
only a small fraction of surveyed users may be involved in
antipiracy and those users may not indicate their intention
because of any concern for legal action against them.
Despite these challenges, a few recent studies have examined the effect of specific antipiracy actions on the behavior of
a particular group of users (i.e.consumers) in a single country
using measurement [1], or survey of users [2] or businesses [3].
All these studies presented a collection of evidences to illustrate that the enforcement of local antipiracy laws succeeded
in reducing the downloading activity of copyrighted content
among their target group of consumers. To our knowledge, the
effect of antipiracy actions on content publishers have not been
13-th IEEE International Conference on Peer-to-Peer Computing
examined, and it is essential because they feed the ecosystem
of online piracy and in some cases gain substantial profit [4].
In this paper, we investigate the effect of antipiracy actions
on the publishers of copyrighted content. To cope with the
challenges in tackling such a broad question, we limit the
scope of our study in two ways as follows: First, we only
examine the effect of two major antipiracy actions: (i) the
closure of Megaupload was a sudden event that was publicized
worldwide, and (ii) the French antipiracy law (Hadopi law)
that was debated, legislated and fully implemented over a
two year period. We intuitively expect these antipiracy actions to have a dominating impact on the behavior of their
corresponding group of users. Therefore, any potential error
in our analysis due to potentially unknown factors should
be relatively small. Second, we only consider the effect of
these two antipiracy actions on the content publishers in the
largest BitTorrent portal, namely the The Pirate Bay (TPB).
Since a significant majority of BitTorrent publishers upload
copyrighted material [5][6], they provide a large population
of publishers that are actively engaged in online piracy and
therefore their reactions offer relevant and meaningful insights
for this study.
One key contribution of this paper is our methodology to
leverage the reaction of BitTorrent publishers for assessing
the effect of selected antipiracy actions. Toward this end, we
capture snapshots of all BitTorrent publishers along with their
uploaded (and downloaded) files through TPB. The timing
of our snapshots are properly aligned to the target antipiracy
actions to increase the likelihood of detecting any measurable
effect even if its impact is short-lived. We use the changes
in the daily number of relevant BitTorrent publishers and
their contribution over the proper time frame as our basic
metrics to assess the effect of each antipiracy action. We show
that this basic metric does not always paint a clear picture
of publisher behavior. Therefore, we deepen our analysis by
grouping publishers based on different criteria to identify
the most likely cause of the observed changes in publishers’
behaviour. These criteria include: (i) level of activity (e.g.,
active vs casual publishers), (ii) publishers’ business profile
(e.g., profit-driven vs altruistic) or (iii) monitoring policies of
their hosting facilities (soft vs strict). Finally, we corroborate
our findings with a few independent sources including Google
trends and other reports to gain more confidence. While there
is not ground truth to validate our findings, we believe that the
number of discovered evidences and their temporal alignment
offer a very convincing explanation for how these selected
antipiracy actions have influenced the behavior of BitTorrent
publishers.
The second key contribution of this paper is to demonstrate
some of the subtleties in identifying the potential effect of an
antipiracy action, and properly relating them to their cause.
These findings of our “detective work” are summarized as
follows:
The closure of Megaupload: Many publishers joined BitTorrent most likely from Megaupload (and other Cyberlockers)
right after its closure. This resulted in an increase in the overall
number of TPB publishers but, surprisingly, had no impact
on their overall publishing rate. This is due to the fact that
major BitTorrent publishers that maintain a private BitTorrent
portal (i.e. a similar business to a Cyberlocker) reduced their
publishing rate in reaction to this event.
French antipiracy Law: The French population have followed the legislation and implementation of the 3 strike law
that targets both consumers and publishers on any copyrighted
content through P2P applications. We show that the first two
steps of the Hadopi law have been very effective in decreasing
the number of casual publishers that as we demonstrate,
are indeed active consumers. However, the number of active
publishers (i.e. uploading more than one content per day on
average) remained stable and they considerably increased their
publishing rate. This reaction is surprising given the reduction
of French consumers for copyrighted content through P2P
applications as reported in [2] and [7]. Our closer examination
revealed that most of top French publishers do not publish any
French content. In fact, the concentration of these publishers in
a particular hosting facility in France (called OVH) appears to
be motivated by the absence of a strict policy for avoiding the
use of BitTorrent on its servers. These professional publishers
are legally savvy and realize that the opportunity to freely
operate from OVH simply outweighs any unlikely antipiracy
action as a result of the Hadopi law.
The rest of this paper is organized as follows: We briefly
describe few prior work in Section II. Section III describes an
overview of our data collection techniques and our datasets.
In Section IV, we present our first case study on the effect of
Megaupload closure. Our second case study on the effect of
French antipiracy law is discussed in Section V. We conclude
the paper in Section VI.
II. R ELATED W ORK
There have been several studies by behavioral scientists on
the motivation of users to engage in online piracy [8], [9],
[10], [11]. These studies typically rely on the collected data
from a small-scale survey (a couple of hundreds user). We are
only aware of two prior measurement studies on the effect of
antipiracy events on illegal file sharing among Internet users.
First, Alcock et al. [1] recently analyzed the impact of the
New Zealand antipiracy law on different applications. They
monitored the traffic of DSL connections for 4000 users at
three different time periods in 2011 and 2012. Their study
demonstrates that the consumption of copyrighted material has
decreased among users and concludes that this is the effect of
the local antipiracy law. Similar to this study, our work relies
on a collection of evidences to draw a conclusion about user
behavior. However, we focus on the behavior of publishers
(rather than consumers) who are clearly engaged in online
piracy.
Second, Lauinger et al. [12] examined the contents of a large
number of uploaded files in eight Cyberlockers to measure the
impact of Megaupload closure on the availability and lifetime
of copyrighted files in other Cyberlockers. They demonstrate
that after Megaupload closure, other Cyberlockers proactively
2
13-th IEEE International Conference on Peer-to-Peer Computing
Crawling Period
Duration (days)
Pubishers (username)
Torrents
Consumers
Downloads
pb10
pb11
pb12
10-09-04 – 10-05-05
27
7.1K
38.2K
27.3M
95.6M
11-10-21 – 11-12-13
54
6.9K
72.0K
25.6M
79.0M
12-01-28 – 12-02-12
16
3.3K
21.0K
5.1M
11.1M
TABLE I
DATASET D ESCRIPTION
increased the filtering of copyrighted material from their
servers, as had been reported in the press [13]. In addition,
in the same study the authors present a qualitative discussion
of the potential impact that the SOPA (US) law [14] could
had achieved if it had been implemented. While this work is
in spirit similar to ours, we examine the behavior of thousands
of BitTorrent publishers (uploading tens of thousands of files).
We are also aware of two prior reports on the effect of
the Hadopi law on French Internet consumers. Preliminary
results of a longitudinal survey of 2K users carried out by the
Hadopi commission indicated a decrease in the download of
copyrighted material [2]. In particular, 72% of the respondents
to this survey who had received a warning, declared that
they had decreased or stopped their activity, and 50% of the
respondent indicated that they have increased the consumption
of legal copyrighted content. The second study [3] analyzed
the data from iTunes record sales by four major labels and
reported 25% increase in the purchase of iTunes music among
French users after the enforcement of the Hadopi law while the
increase in a neighbor country such as Spain was negligible.
Based on these two pieces of evidence, they concluded that the
Hadopi law has been successful in deterring P2P downloads
of copyrighted material.
To the best of our knowledge, our study is the first investigation of the effect of antipiracy actions on the behavior of
publishers of copyrighted content that is based on measuring
a large number of such publishers. More importantly, our
investigation goes beyond obvious metrics and reveals the
impact of other social, economical, and technical factors
through data-driven analysis.
all the indexed content on TPB portal are explicitly uploaded
by a publisher in contrast to the other major portals (e.g.,
Torrentz or IsoHunt) that use crawling techniques to identify
their indexed content. These features make TPB a suitable
venue to capture snapshots of the BitTorrent ecosystem and
conduct our analysis. This section describes a brief overview
on BitTorrent and our measurement methodology as well as
the main characteristics of our collected datasets.
A. Background on The Pirate Bay
TPB is simply a rendezvous point between content publishers and consumers. When a publisher wishes to make a
content available within the BitTorrent ecosystem, its first
step is to generate a unique id known as the infohash and
register the content with one (or multiple) tracker(s). A tracker
keeps track of the IP addresses for a group of peers that
concurrently participate in the delivery of a content (i.e. form
a swarm). A participating peer can be of two types: peers
with a complete copy of a content are known as seeders while
other peers are leechers. Therefore the content publisher is
the first seeder in a swarm. The second step is to advertise
the content by generating a torrent file that provides metainformation for consumers including the IP address of the
associated tracker(s). The publisher uploads the .torrent file
to TPB and possibly other BitTorrent portals. In the case of
TPB, the publisher needs to be registered with the portal and
uses her account (with a specific username) to advertise a
content. TPB creates a separate webpage for each registered
user in which all its published content along with publishing
times are listed. Finally, TPB offers an RSS service where
consumers can subscribe and receive a notification as soon
as a new content becomes available. To download a content,
a consumer typically retrieves the .torrent file from a portal,
extracts the IP address of the tracker and connects to it. The
tracker provides a list of IP addresses for a random subset of
participating peers in the swarm to the new peer so that the
new peer can connect to them and join the swarm.
III. DATA C OLLECTION AND DATASETS
Our objective is to capture multiple snapshots of the BitTorrent ecosystem over time in order to characterize longitudinal
trends in the population and activity of publishers. We use
these trends to assess the impact of antipiracy events on content publishers. Towards this end, we leverage active measurement over The Pirate Bay (TPB) portal using the methodology
and tools that were developed in our earlier study [4]. We focus
on TPB in this study since it is the most popular BitTorrent
portal as reported by scientific studies [15] and Alexa ranking
[16]. In particular, TPB is one of the top-100 most popular
websites in the Internet and receives at least twice (and in
most cases significantly larger) daily visit than any other
BitTorrent portal based on Alexa information. Furthermore,
B. An Overview of the Measurement Methodology
Our measurement tool can capture a rather complete snapshot of all active publishers, their published files and associated consumers within a window of time. To achieve
this goal, our tool subscribes to TPB’s RSS service to get
a notification for any new content that is published on the
3
13-th IEEE International Conference on Peer-to-Peer Computing
portal2 . The RSS feed provides the .torrent file along with the
username of the content publisher. Our tool retrieves the IP
address of the tracker from the .torrent file (or the magnet link)
and immediately connects to it. By connecting to the tracker
immediately after the content is published, we are able to
identify the IP address of the initial seeder (i.e. the publisher’s
location) in many torrents. Our tool periodically connects to
the tracker to retrieve the IP addresses for (typically) 200
randomly-selected participating peers (i.e. consumers) while
respecting the reconnection time imposed by the tracker in
order to avoid being banned. To cope with this limitation,
our tool probes a tracker from eight geographically-distributed
nodes in parallel and captures the IP address of a majority
of consumers. We use MaxMind [17], an IP-to-geo mapping
database, to determine the location of discovered publishers
and consumers. In summary, our captured snapshots contain
the following information for each published torrent on TPB
portal: (i) publisher’s username and IP address, (ii) list of
IP addresses for associated consumers. Further details can be
found at our prior work [4].
Fig. 1. Evolution of Alexa ranking for five popular Cyberlockers over the
last two years (source Alexa).
C. Datasets
offering some income to publishers whose content became
popular [21]. To illustarte the popularity of Megaupload, we
note that Megaupload had 180M registered users and 50M
daily visitors, and stored 12 billions unique files with the
aggregate size of 25 petabytes [19]. We first provide some
info about Megaupload closure and then examine its impact
on BitTorrent publishers.
Using our measurement tool, we have collected three snapshots of TPB system during the past two years. Table I
summarizes the main characteristics of each snapshot including: crawling period, the number of unique publishers,
consumers, torrents (i.e. published files) and downloads for
the three datasets labeled as pb10, pb11 and pb12. Each
dataset was collected over a sufficiently long time such that
any daily or even weekly variations among users and their
activities are captured. The pb11 and pb12 snapshots are
captured shortly before and after the closure of Megaupload
site [18]. Therefore, we use these two snapshots to examine
the impact of Megaupload closure. Moreover, our pb10 and
pb11 snapshots were collected 18 months apart and are used
to investigate the effect of French antipiracy law on French
users.
A. The Closure of Megaupload
On January 19th 2012, the FBI (in coordination with other
agencies across multiple countries) shut down Megaupload
website and arrested their owners on charges of worldwide online piracy that produced $175M unlawful income and caused
$500M loss for the copyright owners [22]. This antipiracy
event had a worldwide coverage. To demonstrate the overall
effect of this well publicized event on the Cyberlockers’
ecosystem, Figure 1 presents the evolution of the Alexa ranking [16] for five popular Cyberlockers over the past two years.
This figure shows two points: (i) Before Megaupload closure,
all of these cyberlockers were either already among the top200 websites in Alexa ranking or their ranking was rapidly
improving until the closure of Megaupload. (ii) After the
closure of Megaupload, the ranking of all Cyberlockers (and
thus their popularity) were rapidly and consistently dropping.
This effect could be due to the adoption of new strategies
by Cyberlockers to actively remove all copyrighted content as
reported by a recent study [12] and in press [13]. In summary,
these evidences confirm that the closure of Megaupload had a
significant impact on all Cyberlockers and possibly all systems
that facilitated illegal sharing of copyrighted content.
IV. E FFECT OF A G LOBAL A NTIPIRACY E VENT
In this section, we investigate how BitTorrent publishers
reacted to a major antipiracy action, the closure of Megaupload
[19]. We focus on this antipiracy action against Megaupload
because it was a major player in illegal sharing of copyrighted
content. Megaupload was the most popular Cyberlocker website. Cyberlockers provide storage service to end users that
enables them to share their online stored content with other
users through a URL. They have quickly become very popular
among users for sharing copyrighted material (e.g., movies,
TV shows, music, etc) through their websites [20]. These
websites became very profitable through posting from ads
and selling premium subscriptions that provide end users with
a better experience (e.g., higher download rate). Moreover,
they even encouraged users to publish interesting content by
B. Effect on BitTorrent Publishers Activity
We rely on our pb11 and pb12 BitTorrent snapshots that
were collected shortly before and after the closure of Megaupload. Given the short time between both snapshots and our
2 Note
that since Feb. 2012, TPB only indexes magnet links instead of
.torrent files. We have accordingly updated our tool to properly operate with
this new indexing strategy.
4
1
0.9
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.4
0.3
0.3
MEGAUPLOAD
SHUTDOWN
400
0.5
0.4
0.2
500
#Publishers
1
0.9
CDF
CDF
13-th IEEE International Conference on Peer-to-Peer Computing
300
200
0.2
100
pb11
pb12
0.1
0
0
100
200
300
400
500
pb11
pb12
0.1
0
0
publisher−per−day
500
1000
1500
2000
0
0
content−per−day
(a) CDF daily publishers
pb11
pb12
367
1334
420 (+14.4%)
1314 (-1.5%)
20
30
40
50
60
70
Fig. 3. Daily number of publishers among those ones collected in our pb12
snapshot during the period 2011-12-01 to 2012-02-12.
Fig. 2. CDF for the number of daily publishers and daily contribution in
pb11 and pb12 datasets.
Avg. daily publishers
Avg. daily contribution
10
DAY (12/01/2011 to 02/12/2012)
(b) CDF daily content
42% of the publishers in pb12 snapshot published their first
file after the closure of Megaupload which suggests that
they most likely joined BitTorrent after this event. Using this
information, we have also determined the aggregate number
of active pb12 publishers in each day during this period as
shown in Figure 3. This figure demonstrates that the number of
daily publishers is relatively stable around 200 until the date of
Megaupload closure and then it rapidly doubles in a few days
once the implications of the event becomes clear to publishers.
These evidences collectively suggest that our observed changes
in the publishers’ demographics and activity between pb11 and
pb12 must be due to the closure of Megaupload.
1) Active vs Casual Publishers: The lack of increase in the
daily number of uploads despite the clear growth in the daily
number of publishers after the Megaupload closure is counterintuitive. To explain this finding, we divide the publishers in
both datasets based on their average daily contribution into
the following three classes: Active publishers that upload more
than 10 contents per day on average, Regular publishers that
upload between one and 10 contents per day, and Casual
publishers that contribute less than one content per day. Table
III shows the average number of publishers per day from each
class and their aggregate daily contributions in snapshots pb11
and pb12. We first analyze the results for pb11 snapshot as
the starting point and later discuss the evolution of each class
between pb11 and pb12.
It is interesting to notice that while the number of casual
publishers in pb11 is roughly three and 20 times larger than
the number of regular and active publishers, respectively, the
overall daily contribution of all three groups is roughly the
same (between 420 to 470 files a day) before the Megaupload
closure. Between pb11 and pb12, the number of casual publishers has increased by 11%, while their contributions remain
unchanged (less than 2%) during this period. The number of
regular publishers has increased by 23% and this has led to a
roughly proportional increase (17%) in their daily contribution.
Finally, the number of active publishers grew by roughly 8%
but their contribution dropped by 21%. In addition, 42% of
casual, 15% of regular and 6% of active publishers in pb12
are newcomers who joined BitTorrent during this period.
In summary, most of the newly arriving BitTorrent publishers after Megaupload closure are casual or regular publishers.
TABLE II
AGGREGATE RESULTS FOR PUBLISHING ACTIVITY IN B IT T ORRENT. T HE
TABLE SHOWS THE AVERAGE DAILY PUBLISHERS AND THE AVERAGE
DAILY UPLOADED CONTENT IN PB 11 AND PB 12. T HE VALUE IN
PARENTHESIS INDICATES THE RELATIVE DIFFERENCE BETWEEN PB 11 AND
PB 12.
target event, and the fact that (to the best of our knowledge)
no other major relevant event occurred during this period,
we are confident that any change in the behavior of BitTorrent publishers is most likely triggered by the closure of
Megaupload. We use two metrics to measure the effect of
Megaupload closure on BitTorrent publishers as follows: (i)
the average daily number of active BitTorrent publishers, and
(ii) the average daily number of discovered uploaded content.
Using these daily average values enables us to compare these
characteristics of publishers across different datasets despite
the differences in dataset durations3 .
Table II presents the average daily number of publishers
and uploaded files for snapshots pb11 and pb12. A more
detailed view of these characteristics is provided in Figures
2(a) and 2(b) that depict the distribution of daily number of
publishers and uploaded files for both snapshots, respectively.
These statistics reveal that the average number of publishers
increased by 14% over 1.5 months whereas their activity
remained roughly unchanged. The observed increase in the
number of publisher (over such a short time) is surprising
and is very likely caused by the migration of publishers from
Megaupload (and other Cyberlockers) to BitTorrent after the
closure since we are not aware of any other event during
this period that can explain such increment. To validate this
observation, we take a closer look at the timing of published
files by individual publishers in the pb12 dataset. To obtain this
information, we have crawled the TPB page of all publishers
in our pb12 dataset and captured the number of files they
uploaded in each day during the 75 day window between
two snapshots (2011-12-01 to 2012-02-12). We observe that
3 We evaluated both metrics for different time windows in pb11 (54 days)
and they remain the same independently of the used window.
5
13-th IEEE International Conference on Peer-to-Peer Computing
n=Avg. content/day
Active (n≥10)
Regular (1≤n<10)
Casual (n<1)
Avg. daily publishers pb11
Avg. daily publishers pb12
Avg. daily publishers difference
% new publishers in pb12 after Megauplaod
Avg. daily contribution pb11
Avg. daily contribution pb12
Avg. daily contribution difference
13.6
14.7
+1.1 (+8.1%)
6.25%
471
374
-97 (-21%)
92.4
113.8
+21.4 (+23.1%)
15.23%
423
510
+87 (+17%)
261
291.5
+30.5 (+11.3%)
42%
440
431
-9 (-2%)
TABLE III
N UMBER OF PUBLISHERS AND DAILY CONTRIBUTION FOR NEXT GROUPS OF PUBLISHERS CLASSIFIED BASED ON THEIR CONTRIBUTION TO THE SYSTEM .
ACTIVE PUBLISHERS ( N≥10 CONTENT / DAY, R EGULAR PUBLISHERS (1≤N<10 CONTENT / DAY ), AND C ASUAL PUBLISHERS ( N<1 CONTENT / DAY )
username
pb11 day cont.
pb11 rank
pb12 day cont.
pb12 rank
Business
URL
scenebalance
TvTeam
exmnova
sceneline
chkm8te
UltraTorrents
FluxXxu
RockSaltS
adultvideotorrents
.BONE.
Black1000
MirrorRu
bigbluesea
eztv
scene4all
107.17
80.9
58.38
53
33.96
24
16.46
15.13
13.06
12.11
11.98
11,65
11.63
11.39
10.29
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
31.06 (-71%)
65.94 (-18%)
35.38 (-39%)
29.93 (-43%)
59.5 (+75%)
6.12 (-74%)
19.6 (+19%)
6.68 (-55%)
2.375 (-82%)
5.25 (-56%)
OUT
OUT
OUT
8 (-29%)
10.56 (+2.56%)
4
1
3
5
2
26
7
24
92
35
OUT
OUT
OUT
18
13
BT Private Portal
BT Private Portal
BT Private Portal
BT Private Portal
Promoting Website
BT Private Portal
Promoting Website
Promoting Website
BT Private Portal
Altruistic
Altruistic
Fake
Altruistic
BT Private Portal
Altruistic
www.scenetime.com
www.torrentday.com
www.69bits.com
www.speed.cd
www.4ufrom.me
www.ultratorrents.com
www.starpix.us
http://jolypic.com/
www.adultvideotorrents.com
TABLE IV
A SUMMARY OF MAIN CHARACTERISTICS ( DAILY CONTRIBUTION RATE AND BUSINESS PROFILE )
OF THE 15 ACTIVE
AND THE CHANGES IN THEIR LEVEL OF PUBLISHING BETWEEN PB 11 AND PB 12.
The overall contribution of three groups were rather balanced
before the Megaupload closure. However, after the closure, the
increase in the contribution of regular publishers is roughly
the same as the decrease in the contribution of active publishers which led to the unchanged overall rate between two
snapshots. This raises the question that “why the relatively
small number of very active publishers have dropped their
publishing rate after the Megaupload closure?” Next, we tackle
this question.
http://eztv.it/
B IT T ORRENT PUBLISHERS IN
PB 11
Websites basically publish content for the sole purpose of
attracting users to their web sites that are often hosting images
services. (iii) Fake Publishers are either antipiracy agencies or
malicious users that inject fake (non-existent) content in order
to warn users of downloading copyrighted material or infect
their computers, respectively. Altruistic Publishers simply publish content to share with others without any expectation of
direct gain.
Table IV shows the following information for the 15 Active
publishers in pb11 in different columns: their usernames, daily
publishing rate, rank in pb11 and pb12, their business profile
and the URL to their web site4 (if applicable). Note that the
Top five publishers in pb11 and pb12 are the same. Among
Active publishers, seven of them are private BitTorrent portals,
three of them are promoting web sites, four are altruistic
publishers, and 1 is fake. The fake accounts are quickly
removed by TPB and two of the altruistic users have also
removed their accounts from TPB. Interestingly, all seven
private portals have significantly reduced their publishing rate
(i.e. their aggregate publishing rate dropped to half from 347
to 178 files per day) whereas other groups of publishers show
mixed reactions.
A plausible explanation for the consistent reaction among
2) Business Profile of Active Publishers: The active publishers that upload more than 10 files a day are in most of
the cases professional publishers behind profitable websites
as we have demonstrated in [4]. Therefore, their wide spread
reaction to (measurably) lower their contributions right after
the Megaupload closure must be related to this event. To
explore this issue, we employ a similar methodology to the
one used in [4] to determine the business profile of all 15
active publishers in pb11. The basic idea in this methodology is to download a few published files by a publisher
and manually inspect whether, where and how a consumer
might be redirected to another web site associated with the
publisher. This methodology broadly divides publisher profiles
into the following categories: (i) BitTorrent Private Portals are
associated with private trackers that offer a better experience
to BitTorrent users for a seeding ratio or a fee. (ii) Promoting
4 Our goal in providing the identity of these publisher is to demonstrate the
fact that many of these publishers are indeed real companies.
6
13-th IEEE International Conference on Peer-to-Peer Computing
100
Assembly and Senate
back Hadopi Bill
Third step.
Expedients sent
out to court
Constitutional Council
accepts amended Bill
40
Second warning
notifications
are sent out
03/03/12
09/10/11
02/27/10
10/10/09
05/23/09
01/10/09
pb11
04/23/11
pb10
20
0
First warning emails
are sent out
National Assembly amends,
supports, and then rejects Hadopi Bill
12/04/10
60
Hadopi Bill
presented
to National
Assembly
07/17/10
80
Date
Fig. 4. Volume search in France of the keyword Hadopi during the period Jan. 2009 to Feb. 2012 according to Google Trends. In addition, we describe the
events associated with major searching peaks.
publishers who manage a BitTorrent private portal is as follows: since the main business model of private portals is very
similar to Megaupload (i.e. facilitating access to copyrighted
material), they decreased their visibility (i.e. footprint) in the
BitTorrent ecosystem to not be viewed as a major player
in order to reduce the likelihood of any antipiracy action
against them. This reaction is actually similar to the one
observed in several Cyberlockers that tried to reduce the
availability of copyrighted content in their portals [13]. Such
a behavior seems to be aligned with the theory in Economics
that punishing a player who performs a non-legitimate activity
generates negative incentive for other players involved in
similar activities [23], [24]. Finally, it is interesting to note that
provided disclaimers in some of the active publishers’ website
confirm that they are clearly aware of copyright infringement
and use the disclaimers to protect themselves against any
potential legal action5 .
western countries have had unsuccessful legislative efforts to
pass a major antipiracy law. For example, the SOPA law in
the US triggered the largest Internet “strike” and was tabled
[14], [26]. The Digital Economy Act in the UK [27] has
also been delayed till 2014 after the appeal by major ISPs
such as British Telecom [28]. The Sinde law in Spain [29] is
going to be ineffective even if it is implemented due to its
bureaucratic process for suing a potential copyright infringing
website [30]. In contrast, there are few countries such as
France, New Zealand [31][1], Korea [32] or Japan [33] that
have passed and implemented antipiracy laws that have been
reported to be (at least partially) successful. Any of these
countries offer a good example for investigating the effect of
such a law. However, we focus on France primarily because
French publishers have a large contribution in pb10, namely
10% of the uploaded content, in the BitTorrent ecosystem
while the contribution of publishers from New Zealand, Japan
or Korea is significantly smaller (<1%). Finally, we are neither
aware of any popular competing technology for legal and
cheap delivery of copyrighted material to users in France (such
as Netflix[34] in the US), nor other antipiracy event happening
in France that could affect the outcome of our analysis.
V. E FFECT OF A L OCAL A NTIPIRACY L AW
In this section, we investigate the effect of a local antipiracy
law in a single country, namely France, on content publishers
that illegally share copyrighted material through BitTorrent.
Toward this end, first we briefly justify our focus on France
and provide the required background on the French antipiracy
law, called the Hadopi law [25]. Afterwards, we examine the
longitudinal trend among publishers as the law was legislated,
approved and implemented.
To limit the number of unknown variables on our investigation, we focus on a country that has a publicized and
properly enforced an antipiracy law. We note that several
Operation of the Hadopi Law: The Hadopi law targets
users that share copyrighted content (i.e. both consumers and
publishers) in Peer-to-Peer (P2P) applications among which
BitTorrent is the most popular one. It is a 3-strikes law that
is implemented as follows: (i) P2P users sharing copyrighted
material are identified by their ISPs and receive a warning
email to stop their illegal activity. (ii) The ISP of the notified
users continues to monitor their activity and if they repeat
their violation during the next 6 months, they will receive a
2nd warning email together with a certified letter. (iii) The
ISPs continue to monitor the notified users for one more year
and if they repeat their violation, the Hadopi commission may
send the violating users to the court. At this stage, a judge
will determine the proper sanction that can be a fine up to
5 An example of such disclaimers is the following: “None of the files shown
here are actually hosted on this server. The links are provided solely by this
site’s users. The administrator of this site (www.69bits.net) cannot be held
responsible for what its users post, or any other actions of its users. You
may not use this site to distribute or download any material when you do not
have the legal rights to do so. It is your own responsibility to adhere to these
terms”.
7
13-th IEEE International Conference on Peer-to-Peer Computing
1
pb10
pb11
487
93
1.4K
156
367 (-25%)
51 (-46%)
1.3K (-6%)
184 (+18%)
0.9
Avg. daily publishers-All
Avg. daily publishers-FR
Avg. daily contribution-All
Avg. daily contribution-FR
0.8
0.7
CDF
0.6
0.5
0.4
TABLE V
C OMPARISON
0.3
OF THE NUMBER OF DAILY PUBLISHERS AND UPLOADED
CONTENT FOR THE ENTIRE B IT T ORRENT (BT) ECOSYSTEM AND IN
F RANCE BETWEEN PB 10 AND PB 11. T HE VALUE IN PARENTHESIS
INDICATES THE NORMALIZED DIFFERENCE FOR EACH METRIC .
0.2
pb10
pb11
0.1
0
0
50
100
150
200
RDU(ratio donwload/upload)
Fig. 5.
Ratio download/upload (RDU ) for casual publishers in France.
suggests that a vast majority of the 1M warnings were actually
received by BitTorrent consumers who must have reacted
by stopping their downloading activity (as demonstrated by
the exponential reduction in the number of warnings sent
out in the subsequent rounds). However, it is uncertain how
French BitTorrent publishers6 , and in particular professional
publishers, reacted to the “Hadopi” law.
1500e and/or the shutdown of their Internet connection for
a period no longer than one month. Further details of the law
can be found in [2][3].
History of the Hadopi Law: The Hadopi bill was first
presented to the French Senate on June 18th 2008. After a
long discussion in the French Assembly and Senate, the law
was amended and passed on June 2009. The last legal step took
place on October 22nd 2009 when the Constitutional Council
finally approved the law. At the end of December 2009, a
committee of experts was nominated to implement the law.
This process took a long time till October 2010 when the
first set of warning emails were sent out. The second round
of notifications occurred at the end of February 2011 (six
months after the first warning). Finally in February 2012, some
expedients were sent to the court as the third strike. It has been
reported that since October of 2010 the number of users that
have received the first and second warning have been 1.15M
and 100K while only 340 expedients have been identified
in the third phase, and 14 have been sent to the court. In
September 2012, the first condemnatory sentence condemned
a user to pay a fine of 150e [7]. Figure 4 depicts a temporal
diagram of the volume of web searches (originated in France)
provided by Google Trends for the keyword “Hadopi” over
which we have specified the time of the above major events
as well as the collection time of pb10 and pb11 snapshots.
The temporal alignment of the pronounced peaks in the search
volume for Hadopi with the time of major events is a clear
indicator that the French population follows this antipiracy
law.
The exponential reduction in the number of warnings sent in
the consecutive rounds indicates that the first two rounds were
the most important since only 340 out of the 1.15M identified
violating users in the 1st round reached the 3rd strike. The
first two rounds took place in October 2010 and February
2011, respectively. Our pb10 snapshot was collected around
April 2010 when the law was passed but still not implemented
and no warning had been sent out whereas the pb11 snapshot
was collected around November 2011, a few months after
the 2nd strike. Therefore the pb10 and pb11 datasets are
suitable to examine the effects of the main two rounds of
the Hadopi on publishers’ behavior. Finally, it is important
to notice that ≥ 99% of BitTorrent users are consumers
(see Table I for number of consumers vs. publishers). This
A. Effect on Publishers Activity
In this subsection, we investigate whether the Hadopi antipiracy law has prompted French BitTorrent (i.e. P2P) publishers to reduce or stop their activity. To tackle this issue, we
examine the average daily number of publishers and uploaded
files among French publishers and compare them with all
BitTorrent publishers (as a reference) in snapshots pb10 and
pb11. The results are summarized in Table V. The drop in
the number of French publishers is roughly twice the drop
among all BitTorrent publishers. However, the daily average
number of uploads for the entire system dropped by 6% while
that measure has increased by 18% among French publishers.
This significant increase in the activity (i.e. uploads) by French
publishers despite the large drop in their number is indeed
surprising.
To further explore this issue, we divide the publishers into
three classes of casual, regular and active publisher based on
their daily average number of uploaded content as we defined
in the previous section. Table VI summarizes the average daily
number of French publishers from each class as well as their
average daily number of uploads in pb10 and pb11. Table VI
shows that (i) the average daily number of casual publishers
and their contributions have both dropped by 57%, (ii) the
average number of regular publishers has dropped by 8% but
their contributions have increased by 17%, (iii) finally, there
are roughly two daily active publishers in both snapshots but
their contribution increased by 144%.
We take a closer look at each one of these trends to
identify their underlying causes. First, our hypothesis is that
those casual publishers (57%) leaving the system are indeed
active consumers who altruistically publish very few content.
Therefore, the Hadopi law has motivated them to stop their
downloading activity which in turn has led to the drop in
their publishing rate as well. We verify this hypothesis by
examining the distribution of the ratio of the number of
6 Those
8
BitTorrent publishers whose location of IP address is in France.
13-th IEEE International Conference on Peer-to-Peer Computing
n= Avg. content/day
Active (n≥10)
Regular (1≤n<10)
Casual (n<1)
Avg. daily publishers pb10
Avg. daily publishers pb11
Avg. daily publishers difference
Avg. daily contribution pb10
Avg. daily contribution pb11
Avg. daily contribution difference
1.96
1.96
0 (0%)
34
83
+49 (+144%)
19.3
17.8
-1.5 (-8%)
66
77
+11 (+17%)
72.6
31.2
-41.4 (-57%)
56
24
-32 (-57%)
TABLE VI
N UMBER OF PUBLISHERS AND DAILY CONTRIBUTION FOR NEXT GROUPS OF PUBLISHERS CLASSIFIED BASED ON THEIR CONTRIBUTION TO THE SYSTEM
IN F RANCE FOR PB 10 AND PB 11 SNAPSHOTS . ACTIVE PUBLISHERS ( N ≥10 CONTENT / DAY, R EGULAR PUBLISHERS (1≤ N <10 CONTENT / DAY ), AND
C ASUAL PUBLISHERS ( N<1 CONTENT / DAY )
downloads to uploads (RDU ) for casual publishers in pb10
and pb11 that is shown in Figure 5. This figure illustrates that
the RDU ratio among casual publishers in pb11 is roughly
an order of magnitude lower than in pb10. This confirms our
observation that most of the departing publishers between our
two snapshots are indeed active consumers with a significant
drop in their publishing activity. Note that these active consumers are likely to be regular Internet users for whom the
sanctions associated with the Hadopi law is considered too
costly (e.g., a fine up to 1500e ).
Second, to uncover the factors that led to the significant
increase in the publishing rate of active users, we take a closer
look at top French publishers. We noticed that one of these
publishers is scenebalance that is the most active one among
all BitTorrent publishers worldwide in pb11 (as shown in
Table IV). Scenebalance is a professional worldwide publisher
injecting more than 100 contents per day into the BitTorrent
ecosystem, most of them from France7 . This raises a couple
of interesting issues as follows:
(i) Since the number of French consumers has rapidly
dropped, these active publishers must be targeting consumers
that are outside France. To explore this issue, we have checked
the published content by top five French publishers in both
pb10 and pb11 snapshots. For pb10 snapshot, we found that
two of the top publishers upload porn content in English,
another one publishes TV series and shows in English, and
the two remaining ones upload only Spanish content. In the
case of pb11, we discovered a similar situation where three
of the top five publishers upload only content in English, and
the remaining two publish Spanish content. We extended this
probe to top 20 French publishers in pb10 and pb11 and
could identify only one publisher who is clearly uploading
content for French consumers (e.g., French content or content
with French subtitles). This investigation confirmed that major
French publishers primarily target worldwide consumers.
(ii) It is then intriguing why these professional publishers
operate from France while their consumers must be mostly
outside France, and there is an enforced antipiracy law that
could affect them. Closer examination of the top 20 French
publishers revealed that more than 80% of them in both pb10
and pb11 snapshots are located at a particular hosting facility
called OVH [35]. In fact, 29 of worldwide top 100 BitTorrent
publishers from pb10 and 25 of them from pb11 were hosted
at OVH. This hosting facility provides professional publishers
with powerful servers to perform their intensive activity of
uploading and serving (i.e. seeding) the large amount of
content they make available through BitTorrent. We contacted
OVH to gain some insight into its popularity among the
professional BitTorrent publishers and learned that OVH does
not proactively monitor the activities of its customers unless
a violation is reported by a third party and the customer
does not cease its “improper” activity [36]. Such a passive
monitoring strategy is unusual as most of the hosting providers
in recent years (e.g., Server Intellects [37]) have adopted strict
monitoring policies to prevent the distribution of copyrighted
material from their servers through P2P applications. These
evidences collectively suggest that the “BitTorrent-friendly”
policy of OVH is much more valuable for publishers than the
cost of any potential antipiracy action against publishers in
France. It is important to note that professional publishers have
major financial interest in publishing copyrighted material [4].
Therefore, they carefully examine any law that might affect
them, take advantage of existing loopholes, and weigh the
likelihood as well as the implications of any legal action
against them. This suggests that even if the Hadopi law intends
to targets publishers, it is much more difficult to deter at least
professional publishers compared to consumers. In a nutshell,
many professional publishers operate from France simply
because OVH’s passive monitoring policy accommodates their
illegal activities.
In summary our results reveal that French antipiracy law
has been quite effective on reducing the number of casual
publishers in BitTorrent who were primarily consumers and
the potential of receiving a fine or temporal loss of Internet
connection as a result of the Hadopi law is considered costly
and thus has a deterrent effect. However, the law has not
succeeded in reducing the publishing rate of copyrighted
material by professional publishers. These publishers seem to
have most of their servers in a particular hosting provider
in France primarily due to the “BitTorrent-friendly” policy of
this provider. The benefits of having access to a facility that
does not monitor the sharing of copyrighted material allows
these businesses to comfortably operate which is clearly more
valuable than the potential risk of any fine that is negligible
compare to their profits.
7 In [4], we demonstrate that usually active publishers upload their content
from different IP addresses that in many cases are located in different
countries.
9
13-th IEEE International Conference on Peer-to-Peer Computing
VI. C ONCLUSIONS
[7] Le Figaro, “Hadopi : un premier internaute condamné.”
[Online]. Available: http://www.lefigaro.fr/hightech/2012/09/13/0100720120913ARTFIG00599-hadopi-un-premier-internaute-condamne.php
[8] T. Ramayah, N. H. Ahmad, L. G. Chin, and L. May-Chiun, “Testing
a causal model of internet piracy behavior among university students,”
European Journal of Scientific Research, 2009.
[9] S. Hinduja, “Trends and patterns among online software pirates,” Ethics
and Inf. Technol., vol. 5, no. 1, Jun. 2003.
[10] W. D. Gunter, G. E. Higgins, and R. E. Gealt, “Pirating youth:
Examining the correlates of digital music piracy among adolescents,”
International Journal of Cyber Criminology, vol. 4, no. 1&2, 2010.
[11] M. W. Kampmann, “Online piracy and consumer affect : to pay or not
to pay,” July 2010. [Online]. Available: http://essay.utwente.nl/60470/
[12] T. Lauinger, M. Szydlowski, W. G. Onarlioglu, K., E. Kirda, and K. C.,
“Clickonomics: Determining the effect of anti-piracy measures for oneclick hosting,” in NDSS’ 13, 2 2013.
[13] PCMAG.com, “After megaupload, storage sites shutter services.”
[Online]. Available: http://www.pcmag.com/article2/0,2817,2399238,
00.asp
[14] A. Bridy, “Copyright policymaking as procedural democratic process:
A discourse-theoretic perspective on acta, sopa, and pipa,” in
Social Science Research Network (SSRN)., 2012. [Online]. Available:
http://ssrn.com/abstract=2042787
[15] C. Zhang, P. Dhungel, D. Wu, and K. Ross, “Unraveling the bittorrent
ecosystem,” IEEE Transactions on Parallel and Distributed Systems,
2010.
[16] http://www.alexa.com.
[17] “MaxMind- GeoIP.” [Online]. Available: http://www.maxmind.com/
app/ip-location
[18] BBC, “Megaupload file-sharing site shut down.” [Online]. Available:
http://www.bbc.co.uk/news/technology-16642369
[19] Wikipedia, http://en.wikipedia.org/wiki/Megaupload.
[20] Envisional, “An estimate of infringing use of the internet.”
[Online]. Available: http://documents.envisional.com/docs/EnvisionalInternet%5FUsage-Jan2011.pdf
[21] Z. Jelveh and K. W. Ross, “Profiting from filesharing: A measurement
study of economic incentives in cyberlockers,” in P2P, 2012, pp. 57–62.
[22] US Department of Justice., “Justice department charges leaders of
megaupload with widespread online copyright infringement.” [Online].
Available:
http://www.fbi.gov/news/pressrel/press-releases/justicedepartment-charges-leaders-of-megaupload-with-widespread-onlinecopyright-infringement
[23] G. S. Becker, “Crime and punishment: An economic approach,” Journal
of Political Economy, vol. 76, 1968.
[24] I. Ehrlich, “Crime, punishment, and the market for offenses,” Journal
of Economic Perspectives, vol. 10, no. 1, 1996.
[25] French Senate, “Haute autorité pour la diffusion des oeuvres et la
protection des droits sur internet,” June’ 2009. [Online]. Available:
http://www.senat.fr/dossier-legislatif/pjl07-405.html
[26] SOPA, http://www.sopastrike.com.
[27] UK Government, “Digital economy act 2010,” 2010. [Online]. Available:
http://www.legislation.gov.uk/ukpga/2010/24/section/1?view=plain
[28] TechWeekEurope, “Britain’s anti-piracy act delayed by cost dispute,”
April 2012. [Online]. Available: http://www.techweekeurope.co.uk/
news/copyright-anti-piracy-act-delayed-75259
[29] Spanish Government, “Ley 2/2011,de economı́a sostenible,” March
2011. [Online]. Available: http://www.boe.es/boe/dias/2011/03/05/pdfs/
BOE-A-2011-4117.pdf
[30] Intereconomia.
[Online].
Available:
http://www.intereconomia.
com/noticias-gaceta/cultura/sgae-%E2%80%9C-valoracion-todos-lossectores-culturales-negativa%E2%80%9D-20120724
[31] Parliamentary Council Office of New Zeland, “Copyright (infringing file
sharing) amendment act 2011,” April 2011. [Online]. Available: http://
www.legislation.govt.nz/act/public/2011/0011/latest/DLM2764312.html
[32] Global Censorship Chokepoints, 2011, http://globalchokepoints.org/
countries/south-korea.
[33] Internet Initiative Japan Inc, “Traffic shifting away from p2p file sharing
to web services,” Internet Infrastructure Review, vol. 8, no. 0, pp. 25–30,
August 2010.
[34] http://www.netflix.com.
[35] http://www.ovh.com.
[36] “Ovh’s terms of service,” http://www.ovh.co.uk/support/termsofservice/.
[37] “Server Intellect Use Policy,” http://www.serverintellect.com/terms/aup.
aspx.
This paper presents a detailed study on how two major
antipiracy actions affect the behavior of publishers in the
largest BitTorrent portal who primarily publish copyrighted
content. In our first case study, we focused on the impact of
the Megaupload closure as a worldwide antipiracy event on
BitTorrent publishers. We showed that the Megaupload closure
triggered an immediate drop in the activity of professional
BitTorrent publishers that are running their own private BitTorrent portals. Furthermore, a group of casual publishers also
migrated to BitTorrent most likely from Megaupload and other
Cyberlockers. Our second case study revealed that the French
Hadopi law was effective in reducing the number of casual
BitTorrent publishers that are actually consumers. However,
it did not have any impact on the activity of professional
publishers from France. The concentration of very active
publishers in a particular hosting facility in France suggests
the popularity of this facility among BitTorrent publishers
that appears to be due to its passive monitoring for copyright
infringement activity. Therefore, legally savvy publishers are
willing to take the chance and operate from France and are not
concerned about a potentially small fine. Our findings provide
a valuable insight about the effect of antipiracy actions on
publishers who are engaged in online piracy and also reveal
the complexity of identifying the affected group of publishers.
While it is impossible to validate our findings, the collection
of all supporting evidences, their temporal alignment and the
dominance of target events suggest that the observed behavior
among publishers are most likely driven by the corresponding
antipiracy events.
VII. ACKNOWLEDGMENTS
This work has been partially supported by the European
Union through the FP7 eCOUSIN (318398) and TREND
(257740) Projects and the ITEA2 TWIRL Project (Call
5-10029), the Spanish Government under the CRAMNET
project (TEC2012-38362-C03-01) and eeCONTENT Project
(TEC2011- 29688-C02-02), the Regional Government of
Madrid through the MEDIANET project (S-2009/TIC-1468),
and the National Science Foundation under Grant IIS0917381.
R EFERENCES
[1] S. Alcock and R. Nelson, “Measuring the impact of the copyright
amendment act on new zealand residential dsl users.” in Proc. of ACM
IMC’12.
[2] T. Meyer, “Graduated response in france: The clash of copyright and the
internet,” Journal of Information Policy, vol. 2, no. 0, 2012. [Online].
Available: http://jip.vmhost.psu.edu/ojs/index.php/jip/article/view/71
[3] B. Danaher, M. D. Smith, R. Telang, and S. Chen, “The effect of
graduated response anti-piracy laws on music sales: Evidence from an
event study in france,” in Social Science Research Network (SSRN).,
2012. [Online]. Available: http://ssrn.com/abstract=1989240
[4] R. Cuevas, M. Kryczka, A. Cuevas, S. Kaune, C. Guerrero, and
R. Rejaie, “Is content publishing in bittorrent altruistic or profit-driven?”
in CoNEXT ’10, 2010.
[5] R. Layton and P. Watters, “Investigation into the extent of infringing
content on bittorrent networks,” Technical Report, 2010.
[6] E. Felten, “Census of files available via bittorrent,” 2010, https://
freedom-to-tinker.com/blog/felten/census-files-available-bittorrent/.
10